News

In response to this research project at least 5 image training datasets were terminated, deactivated, or retracted. The State of AI Report 2020, a Nature magazine feature, and a U.S. Government Accountability Office (GAO) report cited this project's research in what may be turning point in how image training datasets are created and distributed.

September 2021

• September 1: Altinget.dk writes about face recognition datasets in the context of data sharing within Europe (Danish)

August 2021

• August 13: MIT Technology Review writes Deleting unethical data sets isn’t good enough
• August 9: Princeton researchers write about the problem of derivative datasets. Following the citation analysis approach of Exposing.ai, they find that even though datasets like DukeMTMC and MegaFace have been taken down, several datasets derived from them remain accessible and are still in use.

April 2021

• Kate Crawford writes about Exposing.ai in Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence
• April 2: NRC covers our Flickr search engine (Dutch)

January 2021

MegaPixels.cc project becomes Exposing.ai and the New York Times writes: "Here's a Way to Learn if Facial Recognition Systems Used Your Photos.

November 2020

• Nature magazine writes about the ethical issues haunting academic research on facial recognition highlighting MegaPixels research on the Duke MTMC, Brainwash, and MS-Celeb-1M dataset in a growing debate over biometric surveillance technologies

October 2020

• State of AI Report 2020 highlights research from the MegaPixels project about the MS-Celeb-1M face dataseat in their analysis of recent developments in facial recognition. (slide 136)

September 2020

• Sept. 4: Jack Morse at Mashable writes about facial recognition reform, activism, and datasets: People are fighting algorithms for a more just and equitable future. You can, too.

August 2020

• A recent U.S. GAO report on facial recognition cites consent as one of the main privacy issues, referencing research from the MegaPixels project. The full 65 page report: Facial Recognition Technology: Privacy and Accuracy Issues Related to Commercial Uses.

June 2020

• June 11: MegaFace dataset is now decommissioned. University of Washington has ceased distributing the MegaFace dataset citing the challenge has concluded and that maintenance of their platform would be too burdensome.
• The webpage for the Oxford Town Centre dataset has been taken down, and the dataset is apparently no longer being distributed.

October 2019

• Research contribution to New York Times article on "How Photos of Your Kids Are Powering Surveillance Technology"

July 2019

• Our research on Brainwash and MS-Celeb-1M datasets was highlighted in a New York Times article on how featured in New York Times article on how "Facial Recognition Tech Is Growing Stronger, Thanks to Your Face"

June 2019

• June 28: Munich Security Conference launches Transational Security Report including data from MegaPixels on illicit data flows
• June 26: The Atlantic writes about image training datasets "in the wild" and research ethics: Universities Record Students on Campuses for Research by Sidney Fussell
• June 24: Les Echos (FR) writes about MS Celeb dataset and Le mariage explosif de nos données et de l'IA (The explosive combination of our data and AI)
• June 22: La Stampa (Italy) writes about Microsoft's removal of the MS Celeb dataset
• June 15: De Tijd (Belgium) writes about Brainwash head dataset
• June 13: Creator of Duke MTMC dataset apologizes to students recorded for surveillance research to student body and university: "I take full responsibility for my mistakes, and I apologize to all people who were recorded and to Duke for their consequences"
• Jun 12: Duke Chronicle, Duke University's student paper, investigates Duke MTMC dataset, confirms it violated IRB: "A Duke study recorded thousands of students’ faces. Now they’re being used all over the world"
• June 7: Additional coverage of FT's story by BBC, Spiegel.de, IrishTimes, and Gizmodo
• June 6: Financial Times covers the abrupt disappearance of four facial recognition datasets: Microsoft quietly deletes largest public face recognition data set by Madhumita Murgia
• June 2: A person tracking surveillance workshop at CVPR (reid-mct.github.io/2019) has been canceled due to the Duke MTMC dataset no longer being available: "Due to some unforeseen circumstances, the test data has not been available. The multi-target multi-camera tracking and person re-identification challenge is canceled. We sincerely apologize for any inconvenience caused."
• June 2: The Duke MTMC dataset website (vision.cs.duke.edu/DukeMTMC) has abruptly gone blank. An archive from April 18 is still available on the Wayback Machine.
• June 1: The Brainwash face/head dataset has been taken down by its author at exhibits.stanford.edu/data/catalog/sx925dc9385. "This data was removed from access at the request of the depositor."
• June 1: The UCCS dataset page has been updated with a response from the author to clarify that he did not provide any face data to government agencies. Funding was for technology transfer. This site never mentioned that he did provide data to government agencies, only that his work benefited their objectives.

May 2019

• May 31: Semantic Scholar appears to be censoring citations used in this project. Two of the citations linking the Brainwash dataset to research from the National University of Defense Technology (NUDT) in China have disabled. NUDT citation 1, NUDT citation 2, and the original paper show that the NUDT citation has been censored (see the references section on Semantic Scholar pages)
• May 28: The Microsoft Celeb (MS-Celeb-1M) face dataset website is now 404 and all the download links were deactivated. It appears that someone at Microsoft Research has shuttered access to the MS Celeb dataset. Yet it remains available, as of writing this, on Imperial College London's website and on https://msropendata.com/datasets/98fdfc70-85ee-5288-a69f-d859bbe9c737
• May 29, 2019: Stories about the UnConstrained College Students Dataset appeared on Engadget, AP News, New York Times, US News, Daily Dot, Washington Post, MSN, International Association of Privacy Professionals, The Denver Channel, Daily Mail, New York Post, Yahoo! News
• May 27, 2019: Denver Post writes about the UCCS dataset: CU Colorado Springs students secretly photographed for government-backed facial-recognition research
• May 22, 2019: Interview with CS Indy about the UCCS dataset UCCS secretly photographed students to advance facial recognition technology by J. Adrian Stanley

April 2019

• April 20: Washington Post Editorial Board responds to Financial Times article based on data surfaced in MegaPixels project: Opinion | Microsoft worked with a Chinese military university on AI. Does that make sense?
• April 19: Financial Times feature on MegaPixels project: Who's Using Your Face by Madhumita Murgia
• April 19: MegaPixels research cited by report in Financial Times: Western AI researchers partnered with Chinese surveillance firm by Madhumita Murgia
• April 10: Microsoft worked with Chinese military university on artificial intelligence based on data surfaced in MS Celeb dataset

2018

• Aug 22: HRT Transgender dataset on Verge.com: Transgender YouTubers had their videos grabbed to train facial recognition software by James Vincent